is to perform a QR decomposition and then recover all dependent columns. The lcList contains all the linear combos in x. Each item of the list is a list whose first element is the dependent column of the combo:
tmp <- unlist(lapply(lcList, function(x) x[1]))
Thus, my understanding is that after removing tmp there should not be any more linear dependencies in the data. However, you still run a while-loop after the first pass of enumLC where you basically apply enumLC against the data removed of the tmp columns above? A single pass would not be sufficient?
Thanks for developing this amazing package, makes quick model iteration so much simpler.
I was trying to understand the code behind
findLinearComboand I had a few questions about its implementation.I have noticed that the first step of the algorithm:
is to perform a QR decomposition and then recover all dependent columns. The
lcListcontains all the linear combos inx. Each item of the list is a list whose first element is the dependent column of the combo:Thus, my understanding is that after removing
tmpthere should not be any more linear dependencies in the data. However, you still run a while-loop after the first pass ofenumLCwhere you basically applyenumLCagainst the data removed of thetmpcolumns above? A single pass would not be sufficient?Thanks for developing this amazing package, makes quick model iteration so much simpler.